Question on save/restore mfn canonicalization

Question on save/restore mfn canonicalization

[Andres Lagar-Cavilla]

While exploring the paravirt save/restore code path, I noted a 
(seemingly) lack of consistency between where/who canonicalizes 
something and who un-canonicalizes. For example, the guest kernel 
canonicalizes the store and console mfn's in pre_suspend, but 
xc_domain_restore uncanonicalizes them before scheduling back the 
restored guest.
The question is if there is a mandatory reason for this, or is just the 
way the code was written. Can I, e.g, fill the 
pfn_to_mfn_frame_list(_list) entries from "outside" and remove that code 
from post_suspend, or will something break?

Thanks!
Andres

Re: Question on save/restore mfn canonicalization

[John Levon]

On Fri, Sep 28, 2007 at 12:25:28PM -0400, Andres Lagar-Cavilla wrote:

> While exploring the paravirt save/restore code path, I noted a 
> (seemingly) lack of consistency between where/who canonicalizes 
> something and who un-canonicalizes. For example, the guest kernel 
> canonicalizes the store and console mfn's in pre_suspend, but 
> xc_domain_restore uncanonicalizes them before scheduling back the 
> restored guest.
> The question is if there is a mandatory reason for this, or is just the 
> way the code was written. Can I, e.g, fill the 
> pfn_to_mfn_frame_list(_list) entries from "outside" and remove that code 
> from post_suspend, or will something break?

You'll break the ABI. I don't know of a reason why it's so inconsistent.

regards
john

Re: Question on save/restore mfn canonicalization

[Andres Lagar-Cavilla]

Well, when you're in school you don't have to care about breaking ABI's :)
So, the answer is then "no particular reason"?
Andres
John Levon wrote:
> On Fri, Sep 28, 2007 at 12:25:28PM -0400, Andres Lagar-Cavilla wrote:
>
>   
>> While exploring the paravirt save/restore code path, I noted a 
>> (seemingly) lack of consistency between where/who canonicalizes 
>> something and who un-canonicalizes. For example, the guest kernel 
>> canonicalizes the store and console mfn's in pre_suspend, but 
>> xc_domain_restore uncanonicalizes them before scheduling back the 
>> restored guest.
>> The question is if there is a mandatory reason for this, or is just the 
>> way the code was written. Can I, e.g, fill the 
>> pfn_to_mfn_frame_list(_list) entries from "outside" and remove that code 
>> from post_suspend, or will something break?
>>     
>
> You'll break the ABI. I don't know of a reason why it's so inconsistent.
>
> regards
> john
>   

Re: Question on save/restore mfn canonicalization

[Keir Fraser]

Yeah, it's just the way it got written long ago. It isn't as nice as it
could be, but it can't be changed now.

 -- Keir

On 28/9/07 19:55, "Andres Lagar-Cavilla" <andreslc@cs.toronto.edu> wrote:

> Well, when you're in school you don't have to care about breaking ABI's :)
> So, the answer is then "no particular reason"?
> Andres
> John Levon wrote:
>> On Fri, Sep 28, 2007 at 12:25:28PM -0400, Andres Lagar-Cavilla wrote:
>> 
>>   
>>> While exploring the paravirt save/restore code path, I noted a
>>> (seemingly) lack of consistency between where/who canonicalizes
>>> something and who un-canonicalizes. For example, the guest kernel
>>> canonicalizes the store and console mfn's in pre_suspend, but
>>> xc_domain_restore uncanonicalizes them before scheduling back the
>>> restored guest.
>>> The question is if there is a mandatory reason for this, or is just the
>>> way the code was written. Can I, e.g, fill the
>>> pfn_to_mfn_frame_list(_list) entries from "outside" and remove that code
>>> from post_suspend, or will something break?
>>>     
>> 
>> You'll break the ABI. I don't know of a reason why it's so inconsistent.
>> 
>> regards
>> john
>>   
> 
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel